We present a new way to simplify the construction of precise broad-coverage grammars, employing typologicallymotivated, customizable extensions to a language-independent core grammar. Each ‘module ’ represents a salient dimension of cross-linguistic variation, and presents the grammar developer with simple choices that result in automatically generated language-specific software. We illustrate the approach for several phenomena and explore the interdependence of the modules. 1

This paper present amet6 d for improving parsing performance of parsers for HPSG. ThemetL d wasobtjA4A by ext4#6F3 Torisawa's parsingmeti d for HPSG. His parsingmetn dutjL:LF a CFG compiled from a given HPSG-based grammar, and tn parserpredict te possible parse trse wit t CFG. Sincetn amount of unificatni is reduced because oftF6 predict3#A parsing performance is improved. However, we observedtFK tFKM is alimit of speedup bytK metj d. This is duet te fact tac a compiled CFGcannot captt all tl constjKM t in an original HPSG. We add anextK mechanism called array unification t a parser for Torisawa's parsingmeti d and overcometco limitKkLAj The mechanism performspart oft ypeunificat6: and improves precision of parsetrs predict:M wit a small overhead. We observed improvement of parsing speed by afact6 of 2.8wit anexist4# Japanese grammar and 1.7wit anexistLK English grammar. 1

In this paper a speech-to-speech translator from German to English is presented. Beside the traditional processing steps it takes advantage of acoustically detected prosodic phrase boundaries and focus. The prosodic phrase boundaries reduce search space during syntactic parsing and rule out analysis trees during semantic parsing. The prosodic focus faciliates a "shallow" translation based on the best word chain in cases where the deep analysis fails. 1. INTRODUCTION Integration of prosody into a speech-to-speech translator as an additional speech-language interface is a current topic of research. Within the VERBMOBIL project, which aims at translation of spontaneously spoken dialogues, a special experimental system called INTARC was designed which performs translation in an incremental manner. For this purpose time synchronous versions of traditional processing steps such as word recognition, parsing, semantic analysis, and transfer had to be developed. In part completely new algorith...

this article could not have been written. References

Recommended by: